Neuilly-sur-Seine, Île-de-France, France
Hi, I'm François, a computer graphics engineer with a strong interest in real-time rendering and high performance computing. If you are interested in discussing computer graphics, please hit me up on Twitter @frguthmann. https://frguthmann.github.io./
Graphics Productization & Integration Engineer at AMD working at the intersection of real‑time graphics, ML, and tools. I focus on turning ML research projects into shipping products, strengthening FidelityFX/SDK integration, and scaling my impact through technical leadership and mentoring. ML Frame Generation (MLFG): Took an ML‑based frame interpolation project from research prototype to shipping product (DLL + driver override), meeting performance and quality targets across multiple GPUs. Designed its SDK integration, data flow, and memory/shader strategy to minimize hitches, and designed a low‑friction pipeline so new models could be dropped in with minimal code changes. Neural Radiance Cache (NRC): Did some early viability work and API design for NRC, including a DXR path tracer, data pipelines, and shader/ML parity checks. Helped define what a potential FidelityFX NRC product could look like and gathered early partner feedback. Tools & SDK: Continued to maintain and improve key FidelityFX components (e.g., SSSR), fixed multiple stability/perf issues (eg., memory leak) and participated in overall SDK design. Technical leadership & mentoring: Acted as interim tech lead on MLFG for several months, setting priorities, coordinating multiple teams (ML, tools, SDK), onboarding new senior engineers, and keeping a fast feedback loop with research as models evolved. Ran 20+ technical interviews for ML/graphics roles and helped new DevTechs ramp quickly on tools and ISV workflows. ISV & ecosystem advocacy: Presented AMD hardware/software roadmaps and Redstone projects (including MLFG/NRC) to major game studios, led technical Q&A, and fed usability feedback back into engineering and product. Worked with DevTech and ISVs to integrate and debug MLFG across a broad set of titles, contributing to a strong multi‑game launch. I enjoy taking cutting‑edge ML graphics ideas and making them real for game teams so they can ship high‑quality features at scale.
Graphics DevTech and Integration Engineer at AMD, partnering with AAA/AA studios to ship faster, more stable PC graphics. I blend deep profiling, source-level fixes, and tooling/SDK work to turn bottlenecks into measurable wins while sharing playbooks that help teams repeat the results. Performance: Delivered double‑digit FPS gains at 4K on major titles; eliminated stalls by orders of magnitude; achieved double‑digit millisecond savings in volumetrics and multi‑millisecond reductions in ray tracing and GPU passes; improved frame consistency via barrier reordering and draw/PSO sorting. Ray tracing: Identified costly acceleration structure/content patterns and proposed practical fixes; collaborated on changes planned in drivers to reduce unnecessary synchronization and improve dispatch efficiency. FidelityFX/FSR: Guided FSR2/FSR3/FSR4 adoption and quality across multiple releases (motion vectors, pipeline placement, mask usage, HDR/color space, frame generation). Helped remove poor defaults and close integration gaps; built frameworks to evaluate and tune color space and upscaler quality. Tools/SDK: Built a PIX→RGP markers bridge (drop‑in DLL) and a replay tool with shader hot‑reload, enabling reliable profiling workflows; improved SDK samples (SSSR, denoiser, Cauldron PBR correctness), fixed stability/perf issues, and added runtime diagnostics. Influenced profiling tools with occupancy guidance adopted in practice. Knowledge sharing: Led multi‑session GPU pipeline deep‑dives and team knowledge exchanges; authored practical guides (barriers, color spaces, shader replacement, debugging workflows); mentored peers; delivered internal/external talks and a GPUOpen article on occupancy. I care about turning complex GPU problems into clear, repeatable workflows, building tools when they don’t exist, fixing them when they do, and documenting what works so teams can move faster.
- Roles - Principal maintainer of 3DVIA's proprietary WebGL based rendering engine. This engine powers multiple applications running on a wide range of browsers and devices. Our main consumer oriented application is the HomeByMe website visited by hundreds of people everyday. This service allows people to model and visualize their homes in 3D with dozens of assets on screen. Member of the Khronos 3D Commerce working group and the W3C WebGPU community working group, working together with other industry leading companies to shape the future of 3D on the web. https://www.khronos.org/3dcommerce | https://github.com/gpuweb/gpuweb - Work Done - Moved the engine to Physically Based Rendering by adding new shaders, a High Dynamic Range pipeline, moving lights & Image Based Lighting to physical units, updating our live asset editing and offline processing tools, teaching & converting artists to PBR workflows and updating and deploying to production our path tracing service. See our custom shading model https://dassaultsystemes-technology.github.io/EnterprisePBRShadingModel/. Continuously improved the engine by adding multi-light support with improved shadow mapping performance and quality, adding a new bloom post processing and continuously improving CPU and GPU performance, mobile compatibility and code quality. Currently writing a brand new WebGPU backend from scratch.
Integrated Nvidia’s Material Definition Language for complex and multi-layered materials within the web-based 3D engine of Dassault Systèmes. (https://www.nvidia.com/en-us/design-visualization/technologies/material-definition-language/) This was part of an effort to achieve consistency across several engines and required cooperation with other international entities. Tasks achieved include offline material processing (C++), live material interpretation with on-the-fly GLSL code generation (JavaScript) and the creation of an interactive material graph editor (JavaScript).
Quick & light tools development: Interactive, real-time, multiple databases reading cartography using web technologies. VBA scripting to improve workflow and reporting efficiency. Data monitoring, reporting etc.
Store keeping and engineering office.